Force plate calibration WCR 20131104 A force plate converts physical forces in three dimensions (Fx, Fy, Fz) and moments about three axes (Mx, My, Mz) to voltages. The voltages can be measured with a analog-to-digital converter (ADC) and saved. The force plate has six voltage outputs that are supposed to correspond to the six forces and moments. The voltages are labelled Vfx, Vfy, Vfz, Vmx, Vmy, and Vmz. An ideal force plate would have the following relationship between physical forces and moments and voltages:1 πππ₯ = π1 πΉπ₯ (1a) πππ¦ = π2 πΉπ¦ (1b) πππ§ = π3 πΉπ§ (1c) πππ₯ = π4 ππ₯ (1d) πππ¦ = π5 ππ¦ (1e) πππ§ = π6 ππ§ (1f) The quantities S1 through S6 are the sensitivities of the forceplate transducers to forces and moments. Due to unavoidable engineering imperfections, each of the six output voltages is slightly affected by other forces and moments. For example, voltage Vfx, which should be sensitive to Fx and nothing else, may actually be something like the following: πππ₯ = 10πΉπ₯ − 0.06πΉπ¦ + 0.08πΉπ§ − 0.02ππ₯ − 0.04ππ¦ + 0.05ππ§ This “contamination” of the output voltages by unwanted forces and moments is sometimes called crosstalk. The sensitivities, are different for each voltage and each force and moment. The following equations summarize the response of the force plate. πππ₯ = π11 πΉπ₯ + π12 πΉπ¦ + π13 πΉπ§ + π14 ππ₯ + π15 ππ¦ + π16 ππ§ (2a) πππ¦ = π21 πΉπ₯ + π22 πΉπ¦ + π23 πΉπ§ + π24 ππ₯ + π25 ππ¦ + π26 ππ§ (2b) πππ§ = π31 πΉπ₯ + π32 πΉπ¦ + π33 πΉπ§ + π34 ππ₯ + π35 ππ¦ + π36 ππ§ (2c) πππ₯ = π41 πΉπ₯ + π42 πΉπ¦ + π43 πΉπ§ + π44 ππ₯ + π45 ππ¦ + π46 ππ§ (2d) πππ¦ = π51 πΉπ₯ + π52 πΉπ¦ + π53 πΉπ§ + π54 ππ₯ + π55 ππ¦ + π56 ππ§ (2e) πππ§ = π61 πΉπ₯ + π62 πΉπ¦ + π63 πΉπ§ + π64 ππ₯ + π65 ππ¦ + π66 ππ§ (2f) The above set of equations can be written as a single equation using vectors and a matrix: πππ₯ π11 πππ¦ π21 π31 πππ§ = π41 πππ₯ π51 πππ¦ ( πππ§ ) (π61 π12 π22 π32 π42 π52 π62 π13 π23 π33 π43 π53 π63 π14 π24 π34 π44 π54 π64 π15 π25 π35 π45 π55 π65 πΉπ₯ π16 πΉπ¦ π26 π36 πΉπ§ ππ₯ π46 ππ¦ π56 π66 ) ( ππ§ ) (3) or simply π½ = πΊπΏ where V is the vector of voltages, S is the sensitivity matrix, and X is the vector of physical quantities (forces and moments). We assume the force plate voltage outputs have been properly “zeroed”, i.e. adjusted so that the voltage is zero when the force or moment is zero. We also assume the force plate has a linear response. 1 (4) We measure voltages and we wish to recover the physical quantities. In other words, we wish to determine the vector X, based on our measurement of V. Using the rules of matrix arithmetic, we pre-multiply by the inverse of S: πΊ−π π½ = πΊ−π πΊπΏ (5) πΊ−π π½ = π°πΏ (6) πΏ = πΊ−π π½ (7) Force plate manufacturer AMTI provides factory calibration information: the 6x6 sensitivity matrix S and the inverted sensitivity matrix S-1, which AMTI calls B. The elements of sensitivity matrix S have units of V/N (columns 1-3) or V/(N-m) (columns 4-6).2 The elements of inverted sensitivity matrix B have units of N/V (rows 1-3) and N-m/V (rows 4-6). Substituting B for S-1, we can rewrite (7) as follows: πΉπ₯ π΅11 πΉπ¦ π΅21 π΅31 πΉπ§ = ππ₯ π΅41 ππ¦ π΅51 ( ππ§ ) (π΅61 π΅12 π΅22 π΅32 π΅42 π΅52 π΅62 π΅13 π΅23 π΅33 π΅43 π΅53 π΅63 π΅14 π΅24 π΅34 π΅44 π΅54 π΅64 π΅15 π΅25 π΅35 π΅45 π΅55 π΅65 πππ₯ π΅16 πππ¦ π΅26 π΅36 πππ§ π΅46 πππ₯ π΅56 πππ¦ π΅66 ) ( πππ§ ) (8) Equation 8 tells us how to undo the effects of crosstalk and obtain physical quantities (forces and moments) from the measured voltages. Crosstalk: A Simple Example Suppose we measure EMG and EKG simultaneously. The EMG is collected from just below the xiphoid, in order to measure diaphragm activity (breathing). A lead II EKG is also collected (LL-RA). The actual voltages due to EMG and EKG are EMG(t) and EKG(t). However, our measuring electrodes pick up some of the other signal, so our measured voltages are π1 = πΈππΊ − 0.2 ∗ πΈπΎπΊ π2 = πΈπΎπΊ + 0.2 ∗ πΈππΊ We want to extract the true EMG and EKG. We do it by solving the system of 2 equations and 2 unknowns (EMG and EKG). π1 + 0.2 ∗ π2 = πΈππΊ − 0.2 ∗ πΈπΎπΊ + 0.2 ∗ (πΈπΎπΊ + 0.2 ∗ πΈππΊ) π1 + 0.2 ∗ π2 = πΈππΊ − 0.2 ∗ πΈπΎπΊ + 0.2 ∗ πΈπΎπΊ + 0.04 ∗ πΈππΊ π1 + 0.2 ∗ π2 = 1.04 ∗ πΈππΊ (π1 + 0.2 ∗ π2 ) πΈππΊ = = 0.9615π1 + 0.1923π2 1.04 By a similar method we also find that 2 The matrix S provided by AMTI has units of of µV/(Vex-N) (columns 1-3) or µV/(Vex-N-m) (columns 4-6). We multiply the AMTI-specified matrix values by the excitation voltage used (Vex=10V for each channel, in the Biomechanics Lab in Rust 210) and by the voltage amplifier gain (gain=1000 in the Biomechanics Laboratory). The net effect of excitation voltage and amplifier gain is that the elements of S are multiplied by 10,000 to get actual µV/N or µV/(N-m). Alternatively, divide each element of S by 100 to get units of V/N or V/(N-m). The inverted sensitivity matrix B provided by AMTI should be multiplied by 100 to give elements with units of N/V (rows 1-3) and N-m/V (rows 4-6). The text files forceplate[n]_inv_sens_mtx.txt contain the AMTI values multiplied by 100. The AMTI inverted sensitivity matrix (B) is not the exact inverse of the AMTI sensitivity matrix (S). This may be due to rounding. I tried to find a pair of matrices that are more nearly inverse and are the same as S and B to within rounding error of the printed factory values. See finetune_invmatrices.vi for how this was done. The resulting matrices are forceplate4_sens_mtx_corr.txt and forceplate4_inv_sens_mtx_corr.txt. (π2 − 0.2 ∗ π1 ) = −0.1923π1 + 0.9615π2 1.04 We can extract the true EMG and EKG using the equations above. We can also see that these equations can be represented with matrices. If X=(EMG,EKG) and V=(V1,V2), then π π π12 πΈππΊ ( 1 ) = ( 11 )( ) π2 π21 π22 πΈπΎπΊ Where π π12 1.0 −0.2 π = ( 11 )=( ) π21 π22 0.2 1.0 Then we can define B=S-1 and obtain π π12 −1 π1 π π12 −1 π11 π12 πΈππΊ ( 11 ) ( ) = ( 11 ) ( )( ) π2 π21 π22 π21 π22 π21 π22 πΈπΎπΊ πΈπΎπΊ = which simplifies to ( π πΈππΊ ) = ( 11 π21 πΈπΎπΊ π12 −1 π1 ) ( ) π2 π22 We define B to be the inverse of S: B=S-1 . π π12 −1 0.9615 0.1923 π΅ = ( 11 ) =( ) π21 π22 −0.1923 0.9613 Then πΈππΊ 0.9615 0.1923 π1 ( )=( )( ) πΈπΎπΊ −0.1923 0.9613 π2